IRC Logs for #crux-devel Wednesday, 2008-12-17

*** rastajames has joined #crux-devel00:03
rastajamestilman: ping ?00:03
jaegerThere is no tilman, there is only Zuul00:11
* rastajames blinks 00:19
rastajamesjaeger: you played around with ghc00:20
rastajamesin other words you got a Pkgfile i can snap of your hands00:20
jaegernever have, sorry00:28
jaegerused hugs back when I had to do haskell00:28
*** namenlos has joined #crux-devel00:58
*** rastajames has quit IRC01:08
*** sepen has joined #crux-devel01:13
*** sepen has quit IRC02:01
*** DarkNekros has joined #crux-devel02:05
*** sepen has joined #crux-devel02:07
tilmanjaeger: i don't get that Zuul thing02:11
*** namenlos has quit IRC02:28
*** namenlos has joined #crux-devel02:29
nipuLi am the key master03:11
*** namenlos has quit IRC03:55
*** namenlos has joined #crux-devel03:57
*** namenlos has quit IRC04:00
*** namenlos has joined #crux-devel04:07
tilmanreminds me of that tactical sekt song ;)05:02
tilmanthat features quotes from the german dub of that movie ._o05:03
*** namenlos has quit IRC07:01
*** namenlos has joined #crux-devel07:06
*** treach has joined #crux-devel09:17
*** pedja has quit IRC09:19
*** pedja has joined #crux-devel09:27
*** aon2 has joined #crux-devel09:31
treachregarding the little "internal compiler error" problem, I've ran memtest on my laptop, and it doesn't seem to to be caused by bad ram.09:34
jaegersame here, ran it for more than 6 hours last night, around 13 passes09:35
jaegercouldn't build anything with 4.3.2 i686, 4.3.2 i586, or 4.2.2 i686 from 2.409:35
treachthe really weird thing IMO, is that it builds sometimes, and sometimes not.09:36
treachsame compiler, same code, same settings.09:36
treachit got better with jue's gcc package, but I still get it now and then.09:37
sepeninternal compiler... you should crash your laptop to get the internals09:37
jaegerI couldn't even build sometimes... openntpd, kernel, everything I tried failed almost immediately09:37
treachI assume you tried again with the same settings..?09:38
* treach scratches head.09:39
treachit's insane that it works sometimes, and sometimes not.09:39
treachI've no idea why.09:39
treachsome things, like python would fail consistently at the same place though.09:40
jaegerI tried each several times, with and without CFLAGS set, for each of those 3 compilers09:44
jaegerbroke in the same spot every time09:44
treachmmh, what about glibc? Anyone tried changing that?09:46
treach(since 2.9 is out)09:46
jaegerI haven't yet09:53
tilmanthe changelog from 2.8 to 2.9 looked really boring09:54
tilmanit's unlikely this is a glibc issue imo09:54
treachtilman: I was grasping for straws. :/09:55
sepenbinutils or other toolchain related app?09:55
tilmani don't understand the question about binutils10:02
sepenwell gcc, glibc and binutils are toolchain based ports, right?10:02
sepenso I wanted to say that maybe binutils was also a candidate for tests10:02
sepensorry Ill look for changes on it before nothing the next time ;D10:03
treach"Ooops" :p10:23
*** aon has quit IRC11:02
*** aon2 is now known as aon11:02
*** Rotwang has joined #crux-devel11:11
*** namenlos has quit IRC11:29
*** treach has quit IRC11:30
*** Rotwang has quit IRC11:54
*** Rotwang has joined #crux-devel11:55
*** treach has joined #crux-devel12:02
*** pitillo has joined #crux-devel12:09
*** pitillo has quit IRC12:13
*** pitillo has joined #crux-devel12:19
*** jue has joined #crux-devel13:37
jue.oO found it possibly ... :-)13:38
treachthat's good news.13:41
juetreach: you can still reproduce the error?13:41
treachno, actually everything has been working today. I guess I could give ff another try.13:42
treachstrange, really. I ran memtest a few times earlier, and it didn't find anything, so I13:43
treach'm cursing the placement of enter.. and currently ruling out bad ram.13:44
tilmanjue: what do you think is the problem?13:44
tilmantreach: :D13:44
jueI've hurried up my i586 build and updated an very old i586 with the core packages, than compiled on that box all instaled opt ports13:44
jueand than I git this:
juestrfuncs.c:961 is a floating point arithmetics, so I suspect that libgmp might be involved13:47
*** Rotwang has quit IRC13:47
juelibgmp tries to optimize much on the build host, even extensions like mmx, sse etc.13:48
juewell, with a generic libgmp the error is gone13:49
jueif someone want/can test it, please install the libgmp from here:13:50
tilmanlet's hilight nipul with that13:51
tilmannipuL: can you try whether libgmp from fixes your ICEs?13:51
tilmanjue: hehe, see the big-ass warning on
tilman GMP is very often miscompiled!13:51
tilmanmaybe we shuld run 'make check' in the port13:52
jueyes, but I always run a 'make check' with a new version13:52
juethe problem seems to be the host specific optimization13:53
tilmanoh, now i got what you meant13:54
tilmani think :]13:54
jueI've added now a --build=i386-pc-linux-gnu to ./configure13:55
jueafter reading the comprehesive gmp.info13:55
treachhah, just got that error with cups. I'm retrying with the new libgmp.13:57
*** Rotwang has joined #crux-devel13:57
tilmantreach: does it die on cups too at a point where it would try to fiddle with floating point math?13:57
treachprobably, I don't remember the exact error though, but it seemed plausible13:58
treachseems like it works now.14:01
juecool, that's #214:02
tilmanjue: awesome14:02
treachlol, it failed again, but this time it's something else. "No space left on device" :D14:03
treachthanks jue, good catch.14:03
juethanks, but let's wait for another validation14:04
juejaeger: can you test it as well, please ?14:05
juetilman: if you run libgmp's configure you will see something like14:08
jueusing ABI="32"14:08
jue   ....14:08
jue  MPN_PATH=" x86/k7/mmx x86/k7 x86 generic"14:08
juemy new port has only 'x86 generic' here14:09
tilmani guess they don't have codepaths for i686 only?14:10
tilmanie plain i686, without mmx fancyness :)14:11
tilmanor maybe all i686 compatible cpus have mmx, i don't even know :D14:11
treachmmmh, pentium-m has mmx..14:11
tilmanwe could check the libgmp machine code for SSE instructions14:12
treachiirc even some i586 have mmx, but I could be wrong about that14:12
tilmanor SSE2, or whatever14:12
tilmantreach: i think pentium-pro was the first cpu to feature mmx14:12
tilmani have a feeling rehabdoll will correct me in a few seconds14:13
juesure, we can try i686 if it really was the problem14:14
juefor now I used i386, because they wrote that it is the best generic setup14:15
jueand it has to run on my very old K6 ;-)14:16
tilmanthere was a 'pentium mmx' ftr ;D14:16
treachI think the slowest was 133 mmx or something, but that was pretty uncommon.14:16
treach166Mhz was pretty common as mmx14:16
treach(jeeez that was AGES ago) =)14:17
jueusing --build=i686-pc-linux-gnu leads to MPN_PATH=" x86/p6 x86 generic"14:42
jueguess that's what we want14:43
jaegerjue: pretty busy, didn't follow the thread of conversation. you want me to test the libgmp from the link above?14:43
jaegerI can do it later tonight when I'm at home, don't have the p3 laptop here14:43
tilmanjue: that looks good to me :)14:43
juejaeger: yes14:44
juehope it works ...14:44
jaegerI'll try it when I get home and let you know14:48
*** sepen_ has joined #crux-devel15:10
nipuLtilman: my latop is in pieces at the moment15:17
*** jtnl has joined #crux-devel15:44
*** Rotwang has quit IRC15:52
*** Rotwang has joined #crux-devel15:55
*** Rotwang has quit IRC15:57
treachtilman: btw, apvlv (rakt competitor ;) ) works pretty well at least on 32bit.15:58
tilman" its behaviour like Vim. "16:00
tilmansounds cool16:00
treachyeah. :D16:00
*** Rotwang has joined #crux-devel16:00
*** jue has quit IRC16:09
*** Rotwang has quit IRC17:02
nipuLexcept it's written in c++17:04
*** pitillo has quit IRC17:13
*** sepen_ has quit IRC17:14
*** jtnl has quit IRC17:31
*** treach has quit IRC17:58
*** DarkNekros has quit IRC17:58
nipuLwell there goes my leptop18:32
nipuLjust snapped a ribon cable clamp on it :\18:33
nipuLto the hard drive :\\18:34
*** j^2 has quit IRC20:00
*** j^2 has joined #crux-devel20:04
*** j^2 has quit IRC20:06
*** j^2 has joined #crux-devel20:06
jaegermy old inspiron just died for good, it seems. power it on and get memory errors before it even his a boot loader21:59

Generated by 2.11.0 by Marius Gedminas - find it at!